Niko, don't underestimate Gen Four's ability, with suitable prompts to produce realistic "I was there at the scene-here's my personal and emotional reporting" type content (whew, let me catch my breath from such a long sentence!). The looming crisis is one of epidemic uncertainty and deep fakery..
In fact Gen Four could have produced my remarks here, including exclamation marks, humorous asides and warnings about itself.
Certainly, I agree that LLMs can fake a perspective, and write things that are quite funny.
My core argument, though, is that they'd still be faking. They still did not actually "witness" the veritable explosion on the oil rig, and so they can still be called out for their BS.
Another thing I didn't mention is that it's arbitrarily simple to "fool" current AI text-generation algorithms. Adding a few quotes to a story throws them off entirely. We'll have to figure this out and, even if we do, I'm sure that AI writers will fool many people anyway.
If the field of AI was to remain stagnant in it's current state, then we could be reasonably assured that readers will weary of AI generated content in time, and in large part return to human writers.
The problem here of course is that the premise of the above statement is almost certainly false.
Personally, I would vote, if I had a vote, that the entire field of AI be put on hold until such time that we demonstrate that we can clean up the existential threat messes that we've already made. But it's not likely that we are that intelligent.
So it's more likely that we'll keep pushing AI forward as fast as we can until we hit some kind of calamity wall. If true, then sooner or later we won't be able to tell the difference between AI content and human content, and after awhile of that, we'll stop caring. Even today, do you really care who or what wrote this post? It's useful, or it's not, right?
By the way, the field of biology is also likely going to push blindly forward an an ever accelerating pace until it hits some calamity wall too.
At the start of your article, you talk about how journalists can withhold information or provide information in order to maximise the emotional effect on the reader. Do you think it would be possible for AI to achieve this purely through a knowledge and application of neurology / psychology? For example, we know about certain things which often increase levels of dopamine - could AI use this knowledge to move readers emotionally?
Hmmm, thanks for reading and good question. I'm not sure what AI will achieve in 2, 3, or 5 years' time. I suspect we'll all be surprised at its capabilities, as we were with GPT-3 and then GPT-4. Its writing and creativeness will surely improve with time, and writers should probably keep tabs on how it evolves and adapt accordingly.
With so much gloom and panic talk surrounding the future of creative work, this was a very refreshing and illuminating piece. I look forward to reading more of your work!
Niko, don't underestimate Gen Four's ability, with suitable prompts to produce realistic "I was there at the scene-here's my personal and emotional reporting" type content (whew, let me catch my breath from such a long sentence!). The looming crisis is one of epidemic uncertainty and deep fakery..
In fact Gen Four could have produced my remarks here, including exclamation marks, humorous asides and warnings about itself.
But it didn't I assure you.
See what I mean?
Certainly, I agree that LLMs can fake a perspective, and write things that are quite funny.
My core argument, though, is that they'd still be faking. They still did not actually "witness" the veritable explosion on the oil rig, and so they can still be called out for their BS.
Another thing I didn't mention is that it's arbitrarily simple to "fool" current AI text-generation algorithms. Adding a few quotes to a story throws them off entirely. We'll have to figure this out and, even if we do, I'm sure that AI writers will fool many people anyway.
If the field of AI was to remain stagnant in it's current state, then we could be reasonably assured that readers will weary of AI generated content in time, and in large part return to human writers.
The problem here of course is that the premise of the above statement is almost certainly false.
Personally, I would vote, if I had a vote, that the entire field of AI be put on hold until such time that we demonstrate that we can clean up the existential threat messes that we've already made. But it's not likely that we are that intelligent.
So it's more likely that we'll keep pushing AI forward as fast as we can until we hit some kind of calamity wall. If true, then sooner or later we won't be able to tell the difference between AI content and human content, and after awhile of that, we'll stop caring. Even today, do you really care who or what wrote this post? It's useful, or it's not, right?
By the way, the field of biology is also likely going to push blindly forward an an ever accelerating pace until it hits some calamity wall too.
At the start of your article, you talk about how journalists can withhold information or provide information in order to maximise the emotional effect on the reader. Do you think it would be possible for AI to achieve this purely through a knowledge and application of neurology / psychology? For example, we know about certain things which often increase levels of dopamine - could AI use this knowledge to move readers emotionally?
Thank you for a very interesting article :)
Hmmm, thanks for reading and good question. I'm not sure what AI will achieve in 2, 3, or 5 years' time. I suspect we'll all be surprised at its capabilities, as we were with GPT-3 and then GPT-4. Its writing and creativeness will surely improve with time, and writers should probably keep tabs on how it evolves and adapt accordingly.
I couldn’t agree more: I think AI will help writing get de-commodified. And let writers have more creative fun with it!
With so much gloom and panic talk surrounding the future of creative work, this was a very refreshing and illuminating piece. I look forward to reading more of your work!